Summary and Conclusions

Michael Wales - NeedMoreHP@yahoo.com
EEL 5874 - Expert Systems and Knowledge Engineering

Version History:

Document	Date	Description
Summary and Conclusions V1.0	April 18, 2002	Initial Document

Table of Contents

• Overall Summary
• Lessons Learned
• Magnitude of Project

Overall Summary

The project was very successful. I was able to create a very user-friendly interface to an expert system shell that is very complex. The web interface was an optional requirement that I wanted to do, and I put the framework in for it fairly early. The system was created to be very expandable, and very modular. That really helped when I took the initial prototype and adapted it to the web system. I was able to simply change out a few modules and have the system work perfectly.

The system seems to be free of any bugs. When used properly, I don't think it would ever have any problems. The web server application has been the only part of the system that has ever stopped prematurely. Fortunately, the application seems to be much more stable, and I typically have to kill the process to update the knowledge behind it.

The project was very challenging, and I was able to expose myself to a field of Computer Engineering that I had not even touched before. I felt very good that I was able to use my past Software Engineering experience to really expand the application into something that is much more robust. I'm very pleased with how the project came out, and I plan on highlighting it on my web-site for a long time.

Lessons Learned

• Be sure to flush() your output buffer, after you println().
• Java is touted as totally cross-platform, but I've had some strange things happen with my application on different operating systems. The web-server application worked perfectly in Linux (Pentium 100MHz). It would do everything but print the build sheet in WinNT 4.0 (Pentium III 1.0GHz) before hanging. It wouldn't even get past the hardware characterization process in Windows 2000 (Duron 950MHz). But the client applet wouldn't even start in Windows98 (K6-2 500MHz), unless the server was the Linux machine!?!?!
• It is a great advantage, and a huge pain in the butt to develop and test on 4 different computers. If I didn't have the Pentium 100 Linux machine, I probably would have never gotten the ComputerAdvice web-server application working just right. It gave me a great variety of environments to test and program in. I've learned that Windows2000 is the best OS to develop in, while the Linux machine was the best to test the application on due to it's stability. Working on different computers got me into some trouble though when the cable modem quit, and I was unable to download the files to the other computers I needed them at.
• ALWAYS, ALWAYS, ALWAYS test your presentation out on a different computer before giving it. I'm still in disbelief that Java applets can not run correctly from the standard UCF laboratory and classroom computers.

Magnitude of Project

Computer.clp:

Rule Count	37 Rules
Max Number of Rules fired during execution	260
Max Number of facts created during execution	160
Size of file	36 kB
Time spent coding	30 hours

Source Code Modules:

Source File	Lines	Time
CAConnection.java	82 lines	0.5 hours
CAWebClient.java	69 lines	0.5 hours
ComputerAdvice.java	23 lines	1 hour
HardwareCosts.java	240 lines	8 hours
HardwareListCatagory.java	13 lines	0.5 hours
JessInterface.java	161 lines	6 hours
JessInterfaceToWeb.java	200 lines	3 hours
JESSReader.java	37 lines	1 hour
Total	825 lines	20.5 hours